The construction of highways in mountainous regions involves extensive excavation projects. Large-scale excavation can easily disrupt the original topography and geomorphology of the mountains, altering the initial stress state and subsequently inducing slope deformation and failure. The Loess Plateau, an important industrial and agricultural base in China, has a dense transportation network. In recent years, over 100,000 kilometers of roads have been constructed, involving extensive high-fill and deep excavation projects in mountainous areas. Due to the unique properties of loess and the diversity of geological layers, numerous landslides have occurred during construction (Fig. 1). Extensive research indicates that rainfall is the primary factor inducing loess landslides (Ng and Shi 1998, Chang et al. 2021, Dominguez et al. 2021). In addition, excavation during the construction process also affects slope stability (Wang et al. 2014, Sangseom et al. 2017, Meng et al. 2021, Togashi et al. 2022). Landslides during the construction process severely impact project progress, increase construction costs, and pose significant threats to the safety of engineering personnel.
The Loess Plateau exhibits typical stratification (Chen et al. 2024), resulting in the presence of various soil types within the same slope. The significant differences in the properties of these soils, combined with changes in stress states, lead to differential deformation, which is the intrinsic mechanical mechanism of landslides. Silty soil and clay are the most widely distributed soil types in the Loess Plateau. Due to their different origins, they exhibit significant differences in particle size, composition, hydraulic properties, and mechanical characteristics (Kozubal et al. 2024, Sun et al. 2024). However, these two soil types often appear as continuous layers. Once the external stress field changes, they are highly susceptible to differential deformation (Haghighi 2011, Bhat et al. 2013, Musso 2020). Numerous studies indicate that soil landslides often exhibit significant creep characteristics (Sivasithamparam et al. 2015, Chang et al. 2020, Lian et al. 2022, Ismail 2022, Wang et al. 2023, Duan et al. 2024). Compared to rock creep, soil creep deformation is more pronounced at the macroscopic level, resulting in larger deformations (Oliveira et al. 2019). Therefore, numerous fractures can be observed on slopes prior to landslide occurrence. Factors influencing soil creep behavior include not only the inherent properties of the soil but also external conditions such as water content, temperature, and stress (Conte et al. 2010, Sasaki 2015, Yerro et al. 2016, Jeff et al. 2020, Karel 2021). Water content alters the soil's microstructure, affecting its cohesion, internal friction angle, pore water pressure, and viscosity, thereby changing its mechanical behavior (Fletcher et al. 2002, Peng et al. 2022, Li et al. 2023, Duan et al. 2023, Guang et al. 2023). Temperature indirectly affects soil properties by influencing soil water vapor migration (Kong et al. 2021, Xu et al. 2022, Sun et al. 2022). Stress state is a crucial external factor impacting soil creep behavior, encompassing stress history, stress path, and stress corrosion. In the research, factors such as pre-consolidation stress, loading paths, unloading paths, and cumulative damage of the soil were considered (Grammatikopoulou et al. 2008, Zhou et al. 2014, Pei et al. 2017, Kamoun and Bouassida 2018, Yan et al. 2020). During highway excavation, the changes in the soil stress field under unloading conditions are the primary focus (Burland et al. 1977, Mohammadi et al. 2016, Botero et al. 2020, Troncone et al. 2021). The unloading path varies depending on the excavation method. Additionally, the mountain area has a history of numerous ancient landslides, which are prone to reactivation under excavation influence (Yao et al. 2013, Zhu et al. 2022, Wu 2022). In summary, the excavation of slopes in the loess mountainous regions is characterized by diverse strata, climatic variations, and complex geological conditions. These factors determine the varied disaster mechanisms and failure modes, making it difficult to develop a unified landslide prediction model or method. Therefore, current landslide early warning systems primarily use rainfall as the main threshold.
In conclusion, rainfall is the primary factor inducing loess landslides. Thus, research on landslides triggered by highway excavation is closely linked to rainfall, with some studies also considering the effects of excavation unloading and construction vibration loads. The main research methods include numerical simulations, indoor model tests, and field case analyses (Zhang et al. 2020, Wang et al. 2022, Raouf et al. 2024). However, the majority of the Loess Plateau is located in a semi-arid climate zone, with rainfall concentrated between July and September. Extensive engineering practice has shown that landslides do not exclusively occur during the rainy season; they are also likely to happen in winter and spring (Wu et al. 2021, Xian et al. 2022). Therefore, studying the mechanisms and failure modes of non-rainfall-induced landslides is equally significant and can further enrich the landslide research system. As the period of maximum slope disturbance, the construction process is highly prone to landslides. However, there is relatively little focus on landslides during construction, especially research considering the differential unloading creep behavior of soils.
Therefore, this paper uses the reactivation of an ancient landslide in a silty-clay composite stratum during the construction period in the Loess Plateau as a case study. Through field investigations and laboratory tests, the differential creep behavior of silty soil and clay under different unloading paths was analyzed. Numerical simulations were conducted to invert the deformation of the landslide, proposing the corresponding failure mechanism and disaster evolution model. Based on the soil creep model, a stress attenuation-based landslide prediction model was developed. The research results can provide technical reference for related engineering projects and offer guidance for future studies.